This is an earlier experiment. Read the amendments below before treating its findings as current.
The question
Craft reaching a lane's end restart at its start. Are the wing's losses, about 100 per 12 runs at every wind, a side effect of that? Where and when do collisions happen?
What was tested
No new flights: every collision in 72 wind and 12 calm runs was classified from its last 7.5 s, located and timed.
What was found then
Recent restarts were 9–32 % of wind-run collisions per run; 84 % involved neither restart nor launch, mostly between different lanes and early (median 45 s). In calm the wing lost 532 per 12 runs, 5× more than in any wind, read then as wind spreading traffic out at crossings. Only a third of calm collisions were within 100 m of a drawn crossing; most were between perpendicular lanes, away from it. After three minutes calm runs kept colliding (198 more) while wind runs almost stopped, read then as wind ending the process.
What we learned later
The label was corrected from 'genuine' to 'mixed' (at one wind, re-placed craft were a third of collisions per run), and 'wind spreads traffic out' was restated as an untested idea.
research/envelope.md:2512
Narrowed: in the first three minutes, wind runs collide 8–20 % less than calm, not 'at the same rate'. Whether wind ends the process was never tested; the three-minute split was chosen after seeing the data.
research/envelope.md:2572
Every stage-1 wind run used one wind direction. In stage 5, the loss reduction did not reliably carry over to wind from 70°.
research/envelope.md:8592-8600
What changed
Nothing.
Numbers and source references
- 72 / 12
- wind and calm runs whose collisions were classifiedresearch/envelope.md:2512
- 7.5 s
- position history before each impact used for classificationresearch/envelope.md:2512
- nine
- drawn lane crossings on the gridresearch/envelope.md:2572
- 9–32 %
- mean per-run share of wind-run collisions that were recent re-placementsresearch/envelope.md:2512
- 84 %
- wind-run collisions with neither recent re-placement nor launchresearch/envelope.md:2512
- 45 s
- median impact time of those collisionsresearch/envelope.md:2512
- 532
- wing craft lost per 12 calm runs (600 craft each, 30 min)research/envelope.md:2512
- 5×
- calm losses relative to the 98–120 at wind 6–7.5 m/sresearch/envelope.md:2512
- 33 % of 249
- calm collisions (no recent re-placement or launch) within 100 m of a drawn crossingresearch/envelope.md:2572
- 13.5 %
- share of fleet time within 100 m of a crossingresearch/envelope.md:2572
- 198
- calm such collisions after 180 s (12 runs)research/envelope.md:2572
- 1–4
- such collisions after 180 s per wind armresearch/envelope.md:2572